2020 (modified: 06 Oct 2021)ICML 2020Readers: Everyone
Abstract:Recent advances in deep learning theory have evoked the study of generalizability across different local minima of deep neural networks (DNNs). While current work focused on either discovering prop...